Questions:

• Apply systematic risk management rules to reduce impact from high volatility market environment.
• Goal to improve risk properties of index more than sacrifice w.r.t. annual returns.
• Is there a generic set of rules that consistently improve the index Sharpe ratio?

Approach:

• Implement four risk thresholds with corresponding portfolio overlay weights.
• Scan a parameter grid w.r.t. the risk thresholds and weights for a set of alpha AiLA indices.
• Try to identify a parameter set that consistently improve the index Sharpe ratio.

• The risk management parameters for a given AiLA index are defined w.r.t. the annualized sample std.dev. of index returns, i.e. normalizing the performance w.r.t. the typical annual return size of the index.
• Annualized risk can differ significantly, since some indices are constructed with a specific vol. target and others with varying index vol.
• In case the cumulative index PnL cross any of the threshold values, the corresponding weight is applied thereafter to the index asset allocations, i.e. this refers to crossed from below (above) for the portfolio target and soft loss reset (soft loss and hard loss) thresholds.
• In the exercise all weights are reset to 1.0 annually.
• The risk-overlay is applied to each index for 163 grid points,
  • 𝐶_𝑝𝑡 : (0.25, 0.50, 1.00), 𝑊_𝑝𝑡 : (0.25, 0.50, 0.75)
  • 𝐶_𝑠𝑙 : (0.10, 0.25, 0.50), 𝑊_𝑠𝑙 : (0.25, 0.50, 0.75)
  • 𝐶_ℎ𝑙 : (0.10, 0.25, 0.50), 𝑊_ℎ𝑙 = 0
  • 𝐶_𝑠𝑙𝑟 = 0, 𝑊_𝑠𝑙𝑟 = 1.00

• The risk management overlay was applied to 20 AiLA indices, for each point of the parameter grid.
• For 8 out of the 20 indices less than 10% of the points corresponded to an improvement w.r.t. the Sharpe ratio of the original index.
• From the remaining 12 indices, 3 indices where ≳50% of the points corresponded to an improved Sharpe ratio w.r.t. the original index, were used to select favorable parameter settings in-sample.
• The remaining 9 indices were then used to investigate if the selected parameter settings show indications of being generically favorable out-of-sample.

• Using the 3 in-sample indices parameter points were filtered out based on the following criteria,
  • Nr of days with allocation > 50% w.r.t. original index.
  • SR(risk-overlay) > SR(original)
  • Rank(SR) among all points > 2/3 x median(SR)
• This resulted in two favored parameter grid points,
  • (𝐶_𝑝𝑡 , 𝑊_𝑝𝑡 , 𝐶_𝑠𝑙 ∙ 𝐶_𝑝𝑡 , 𝑊_𝑠𝑙 , 𝐶_ℎ𝑙 ∙ 𝐶_𝑝𝑡)
  • Point 61: (0.50, 0.25, -0.05, 0.75, -0.125)
  • Point 79: (0.50, 0.50, -0.05, 0.75, -0.125)
• Ideally these parameter settings could yield an improved as well as a high ranked Sharpe ratio also for other indices?
• Both the parameter sets result in 2/9 indices with improved SR and 3(4)/9 indices with rank(SR) above the 81 median value.
• The Sharpe ratio improvement is often generally common or rare for a given index, i.e. difficult to interpret. However, together with the ranked results, no indication of a general SR improvement by the in-sample grid points was observed.